Identification of microRNAs regulating Escherichia coli F18 infection in Meishan weaned piglets

Reviewer summary

This paper suggests that several miRNAs, notably mir-218-3p, are involved in E. Coli resistance in Meishan piglets, via targeting DLG5 among others. The finding is interesting but the ms. needs far better description of methods, and better analysis and presentation of the data. The fact that the experiment was based on only 3 animals per group is troubling!

Reviewer recommendations to authors

1/Since most readers may be unfamiliar with Chinese pig strains, I suggest that you discuss in more detail why you chose Meishan piglets as a model system.

Author’s response: Thanks for your suggestion. We added some description about Meishan pigs in discuss section: “Taihu pig is one breed with the highest reproductive traits in Chinese and foreign pigs around the world. Thereinto, Meishan is the most prominent representative population of Taihu pigs and has the advantage of higher litter size, more delicious meat, etc. Furthermore, compared to commercial western pig breeds, the Chinese Meishan pigs exhibit not only the higher litter size but also an increased physiological maturity which is correlated with the piglet survival rate before and at birth.”

2/The Methods need to describe much more clearly how you challenged the piglets and what time lag between your challenge and the tissue extraction for miRNA levels. If I am correct, you have no way of telling in advance which piglets will be resistant? Sounds like you isolated duodenum during active infectious phase, which might complicate the pattern in the susceptible piglets? That should be discussed as possibly affecting your results.

Author’s response: According to your requirement, we added detailed description about the method of challenging the piglets in our manuscript. As following:

“We selected three litters of weaning piglets at 35 days of age, 12 piglets per litter, with almost same birth weight and weaning weight. Twelve piglets per litter were randomly divided into two groups: the control group (two piglets) and the experimental group (ten piglets). Each piglet was housed individually in separate pens. They were fed ad libitum with a commercial-type compound feed for weaned piglets containing 21.7% crude protein, without antimicrobial additives and organic acids. Beginning at day 3 post-weaning, experimental piglets were challenged with a daily dose of 4.6×10 ⁸ CFU of E. coli F18 strain once a day for up to 10 days, or until they showed diarrhea. No additional food was given and we ensured that piglets ate all food before the challenge experiment. Throughout the experiment, fecal shedding of the inoculated bacteria was monitored by daily fecal sampling and feces consistency was scored using the parameters “normal”, “pasty” and “watery”. Only piglets with watery feces were considered as diarrheic. The intestinal tracts of the diarrheic pigs were used to carry out a series of experiments, such as E. coli F18 bacteria counting, histopathological detection and adhesion test of the pathogens to the epithelial cells of small intestine in vitro [48].”

For the identification and selection of E. coli F18-resistant and –susceptible piglets, we mainly used the following method:

“Nine piglets showing “watery diarrhea” were defined as the “diarrhea group” and eight normal piglets as the “normal group”. After slaughter, we took intestinal tissues (duodenum, jejunum and ileum) to detect bacterial numbers. In all piglets detected by the binding assay, we strictly identified piglets displaying no adherence with F18-expressing fimbriae of the standard ETEC strain as E. coli F18-resistant individuals (Additional file 6A). In contrast, piglets displaying a large amount of adherence were identified as E. coli F18-susceptible individuals (Additional file 6B and 6C).”

Additional file 6. Adhesion test for intestinal epithelial cells for E. coli F18-resistant and -sensitive piglets. The adhesion of Escherichia coli F18 to intestinal epithelial cells in Meishan piglets, A represents F18-resistant piglets displaying no adherence with F18-expressing fimbriae of the standard ETEC strain; B represents F18ab-susceptible piglets displaying a large amount of adherence with F18ab-expressing fimbriae of the standard ETEC strain, C represents F18ac-susceptible piglets displaying a large amount of adherence with F18ac-expressing fimbriae of the standard ETEC strain. Photos were taken with an oil immersion lens at 1000× magnification.

48. Liu L, Wang J, Zhao QH, Zi C, Wu ZC, Su XM, et al. Genetic variation in exon 10 of the BPI gene is associated with Escherichia coli F18 susceptibility in Sutai piglets. Gene. 2013;523:70–75.

3/When only 3 animals per group are studied, there is a very high risk of false-positive findings, even with 2-fold elevations as seen here. I strongly urge you to carry out permutation analysis, i.e. take the set of 6 animals and randomly assort them to two groups (n=3 in each group) in all combinations. Do you find a similar number of 2-fold changes in miRNAs in all permutations, or only in the case where all 3 samples are susceptible and the other 3 are resistant? The traditional p-values as done here are not sufficient to ensure that the data are robust.

Author's response: In our study, we performed a comparative microRNA transcriptome study on duodenum tissues between Meishan sensitive group (n=3) and resistant group (n=3) using Illumina Solexa sequencing technology. For high-throughput sequencing, generally speaking, 3 repeats meet the requirements of sequencing analysis. Besides, most studies also have 3 biology repeat in sample processing (Bai et al., 2014; Chen et al., 2012; Tao et al., 2013).

Bai Y, et al. (2014) A comprehensive microRNA expression profile of the backfat tissue from castrated and intact full-sib pair male pigs. BMC Genomics 2014, 15:47.

Chen C, et al. (2012) Solexa Sequencing Identification of Conserved and Novel microRNAs in Backfat of Large White and Chinese Meishan Pigs. PLoS ONE 7(2): e31426.

Tao X, Xu Z (2013) MicroRNA Transcriptome in Swine Small Intestine during Weaning Stress. PLoS ONE 8(11): e79343.

About the identification of differential miRNAs, we performed the comparison of the known miRNA expression between two groups (sensitive and resistant group) to find out the differentially expressed miRNAs. The expression of miRNA was shown in two samples by plotting Log2-ratio figure and Scatter Plot. Firstly, we normalized the expression of miRNA in two samples (sensitive and resistant group) to get the expression of transcript per million. When the normalized expression of a certain miRNA was zero between two samples, we revised its expression value to 0.01. While if the normalized expression of a certain miRNA was lower than 1, further differential expression analysis was conducted without this miRNA. Normalized expression (NE)=Actual miRNA count/Total count of clean reads×1000000. Moreover, we calculated the fold-change and P-value from the normalized expression. Then generate the log2ratio plot and scatter plot. Fold_change=log2(resistant group/sensitive group).

4/I would strongly urge you to examine another cohort of piglets and at least measure mir-218-3p and other altered miRNAs, to make sure that the changes are reproducible.

Author’s response: Many thanks for your suggestion. In our study, we obtained some differential expression miRNAs related to E. coli F18 by Solexa sequencing and further verification was performed between F18-sensitive and resistant piglets by Stem-loop RT-qPCR. However, your advice is very reasonable. In the next step work, we will detect the expression in different populations of piglets and analyze the function of miRNAs and target genes by knockdown or overexpression at the cellular level. If we achieve some breakthrough results, we hope to continue to get your guide.

5/You appear to have isolated duodenum and jejunum, and measured target mRNAs in both tissues, but did not measure jejunum for miRNAs?? Why not? Especially, it would help to measure mir-218-3p and other altered miRNAs in jejunum.

Author’s response: Escherichia coli (E. coli) are a group of gram negative flagellated bacteria that normally reside and multipl\y in the intestinal tract of all animals. Veterinary pathology experiments demonstrated that duodenum and jejunum are main place where E. coli F18 strain colonizes and replicates. The jejunum indeed can be used as samples for E.coli F18 adhesion test. In previous studies, duodenum was used for E.coli F18 adhesion and high-throughput sequencing (Bao, et al., 2012; Wu et al., 2016), in view of combination the data of previous high-throughput sequencing, we still choose the duodenum in this miRNA sequencing. However, we have isolated duodenum and jejunum for systematic qPCR validation.

Wu ZC, Liu Y, Dong WH, et al. CD14 in the TLRs signaling pathway is associated with the resistance to E. coli F18 in Chinese domestic weaned piglets. Scientific Reports, 2016, 6:24611.

Bao WB, Ye L, Pan ZY, et al. Microarray analysis of differential gene expression in sensitive and resistant pig to Escherichia coli F18[J]. Animal genetics, 2012, 43(5): 525–534.

6/Some primary data are alluded to but should be explicitly presented. For example, you show p-values for correlation coefficients (r =), but you should show the r values themselves directly as well. In Table 1, with only 6 samples, you have room to display individual sample values and SDs in addition to the summary ratios and p-values. This would allow us to see if you have any high variability or outright outlier values in your data, which is crucial when there are only 3 samples per group.

Author’s response: Thanks for your comments. We revised the Table 1 with adding the mean values and SDs. Table 1 shows the differential expression between E. coli F18-resistant group and E. coli F18-sensitive group by qRT-PCR. However, the comparison of miR-SEQ and qRT-PCR has been shown in Fig. 5.

7/The authors should ideally demonstrate that mir-218-3p and DLG5 are both expressed in gut epithelial cells, which is not a given since they carried out measurements on entire duodenal tissue.

Author’s response: In our manuscript, we performed a comparative miRNA sequencing of duodenal tissues between E. coli F18-resistant group and E. coli F18-sensitive group, and then we screened out some differential expression miRNAs including mir-218-3p. For qRT-PCR detection in duodenal tissues, our original purpose is to verify the result of miRNA sequencing. Meanwhile, we could further analyze the correlation of mir-218-3p and DLG5 expression.

However, the expert’s opinion is very reasonable. To analyze the function of DLG5 gene, we performed some on-going studies in gut epithelial cells. Actually, it extremely difficult to obtain pure epithelial cells (possibly include other intestinal cells). To avoid other intestinal cells, we strictly conducted the following sampling process: (1) after slaughter, the duodenum was incubated on ice for 1 h. Then, pre-cooling PBS (PMSF, NaN3) mixture washed the duodenum and exposed the intestinal inner-wall. (2) We scraped the intestinal mucosa with the slides, PBS washed it and centrifuged for 10 min at 200×g. This process is repeated 4 times until the clean precipitation. (3) We added NaHCO ₃ (NaN ₃ , PMSF) mixture and grinded 40 times, then centrifuged for 10 min at 450×g. (4) we proceeded a hung heavy precipitation with separation buffer (EDTA, NaN ₃ , PMSF), grinded 20 times, then centrifuged for 10 min at 300×g. This process is repeated 4 times. (5) Once again, we proceeded a hung heavy precipitation with Mg ²⁺ buffer (EDTA, NaN ₃ , PMSF), grinded 20 times, stewing at 4 °C for 30–60 min. (6) The supernate fluid was filtered by lanoline to remove the nucleus, then centrifuged for 10 min at 450×g. (7) Finally, proceeded a hung heavy precipitation with Final buffer (KH ₂ PO ₄ , Sorbitol, NaCl, NaN ₃ ) and stored at −80 °C until further use. Meanwhile we further confirmed that the sample was only epithelial cells by microscopic examination, see below:

Minor issues

1/You say you “randomly selected” 15 miRNAs to verify by RT-PCR, but the list does not look random.

Author’s response: Thanks for your advice. We have deleted the word “randomly” in our manuscript.

2/The Discussion section talks about some methodological points that belong in Methods.

Author’s response: About some methodological points, such as bacterial infection experiment, we emphasized the three key points to improve the effectiveness and feasibility of our piglet diarrhea model using an artificial challenge experiment.

3/The paper talks about “target genes” which are really target mRNAs.

Author’s response: “target genes” indeed seems a little absolute. We revised “target genes” as “potential target mRNAs”.

Amended comments

The manuscript is improved but I am rather confused and worried by the findings presented. They show that mir-218-3p is down-regulated in sensitive tissues by RT-PCR (Table 1), and DLG5 is also down-regulated in the same tissue (fig. 8), yet they assert that the two are strongly NEGATIVELY correlated (fig. 9). That does not seem right to me! If the two are indeed negatively correlated [across both sensitive and resistant tissues], then one of the two should be UP-regulated in sensitive tissues. It seems that there is a fundamental discrepancy between sequencing data (fig. 5) which shows that mir-218 is UP-regulated in sensitive tissues and the RT-PCR data (Table 1) which shows a significant DOWN-regulation in the same tissue, thus not a validation but instead in opposition. Something is not right, and it is crucial to clarify these before publication.

Author’s response: Many thanks for your pointing out mistakes. We thought the order of miR-name (from Table 1, as follow) was consistent with the detected miRNA (from Original data, as follow). We mistakenly copied original data into Table 1. We felt very ashamed for such a mistake. We revised the Table 1 according to the original data.

* p < 0.05.

Reviewer summary

The manuscript described a study profiling miRNAs in Chinese domestic weaned piglets challenged with E. coli F18 strain which is know to cause porcine post-weaning diarrhea. Two small RNA libraries were constructed using 3 pooled samples of F18-resistant groups and 3 pooled samples of F18-sensitive groups, and sequenced by the Illumina deep sequencing platform. Fifteen upregulated and 9 downregulated miRNAs were identified between the two groups of piglets, and their potential mRNA target genes were identified using results in the Miranda database. The expression of some of the miRNAs and their target genes were experimentally validated using PCR assays. GO enrichment and KEGG pathway analyses were performed on the target genes to assess biological relevance and significance of the results. The overall design of the study and the profiling experiments are sound. The results of differentially expressed miRNAs that are potentially responsive to F18 infection may be valuable for future studies and to practitioners in a focused area. The overall approach taken is conventional and is not novel.

Reviewer recommendations to authors

Two major areas of improvement can be introduced to improve the quality and expand the scope of the study. The first is to identify novel miRNAs using the sequencing data. This can be done using many published methods, e.g., miRDeep. (Or write to me, weixiong.zhang@wustl.edu and I am happy to provide our miRvial tool for miRNA prediction.) The novel miRNAs that are potentially specifically responsive to F18 infection may provide deep insights to miRNA gene regulation. The second area of improvement is to introduce mRNA profiling using RNA-seq. Ideally such data can be gathered using the same total RNA that was used to profile miRNAs. Profiling of mRNA gene expression can be integrated with the miRNA target information to paint a genome-wide picture of miRNA-mRNA regulation. Another possible improvement is not to pool the 3 samples into one library, but rather separate them using barcodes and profile them using multiplexing sequencing. The new data can provide some statistical power in calling differentially expressed miRNAs. In light of these possible improvements, which can be easily incorporated, the current study seemed rather rudimentary.

Author’s response: Many thanks for your valuable suggestions. Firstly, for novel miRNAs, we added the identification of novel miRNAs by mireap software. As following: “To identify novel miRNA genes among the unannotated sequences in our libraries, we employed the mireap program (https://sourceforge.net/projects/mireap/), which processes high-throughput sequencing data sets.” was placed in th8e section “Sequencing data analysis” of “Methods”. Finally, we identified 681 novel miRNAs, as shown in added “Additional file 2”.

Secondly, your advice on mRNA profiling is very reasonable. In our manuscript, we have considered the combination of mRNA and miRNA. In previous studies, we have obtained differential expression genes (DEGs) between Meishan F18-resistant and -sensitive groups (Samples are in complete accord with this miRNA study) by mRNA transcriptome sequencing (data are available at NCBI’s SRA, PRJNA271310). In this study, we also found some differential miRNAs and their potential target genes were predicted, so we further screened out important target genes based on previous DEGs. Moreover, about the samples for sequencing, we did not pool the 3 samples into one library, but six small RNA libraries were constructed from resistant (R1, R2, R3) and sensitive (S1, S2, S3) piglets to E. coli F18, respectively. We will systematically perform an in-depth study of miRNAs and target genes in future, and then hope to continue to get your guidance.

Minor issues

The quality of some of the figures, e.g., Fig 6, should be improved. It’s difficult to see the content of Fig 6 and Fig 7.

Author’s response: According to your requirements, we have improved the resolution of all figures to 500 dpi.

Amended comments

The authors made their effort to answer the reviewers’ questions and comments. However, the additional work they have done didn’t seem to improve the quality of the work. In particular, the authors added the result on novel miRNAs they could predict using an off-the-shelf method on their sequencing data. However, they didn’t integrate the result to help achieve their goal of identifying miRNA gene regulators in the process of E. coli infection. So the additional work of novel miRNA prediction is completely useless to the current study and should be removed if they indeed didn’t want to add any functional information of the novel miRNAs. Second, while in their responses to reviewers’ comments they mentioned that they used their previous mRNA data in the revision, no where in the new manuscript such results or description can be found. In their responses, they indicated that they in fact would like to defer integration of miRNA and mRNA data to a future study. In short, the revision is essentially the same as the original submission, and as such I don’t believe it meet the standard of the journal.

Author’s response: Many thanks for your comments. Referring to some literature, most studies mainly focused on the predicted novel miRNAs, its free energy and stem-loop structure. Therefore, based on the predicted novel miRNAs, we further analyzed the typical stem-loop structure and free energy. Revision in manuscript as following:

In the section “ Sequencing data analysis ” form “ Methods ”, we added “The RNAfold software in the ViennaRNA Package 2.0 [20] was also used to predict the typical secondary structures of the miRNA precursors.” In the “ Differential miRNAs in E. coli F18-sensitive and -resistant Meishan piglets ” from “ Results ”, we added “In addition, 681 potential novel miRNA candidates were obtained from RG and SG libraries. These pre-miRNAs possessed a typical stem-loop structure and free energy ranging from −64.8 Kcal/mol to −18.2 Kcal/mol (Additional file 3). The folding (free energy>50.0 Kcal/mol) are shown in Additional file 4.

Additional file 4: Partial secondary structure of novel microRNAs. Folding secondary structure of novel microRNAs and flanking sequences was predicted by RNAfold. The entire sequence represents pre-miRNAs.

In our study, the original aim is to identify some known miRNAs related to E. coli F18 infection, which is the main point and meaning of this manuscript. As you said, we simply predicted the novel miRNAs and not analyze their function. It is nearly impossible for analyzing the function of all novel miRNAs, and we think that these novel miRNAs probably provide some database information on pig miRNAs, which aimed to provide useful information for future study of other researchers. Because you are an authoritative expert in this field, our study is relatively preliminary and we hope to get your understanding.

About the integration of miRNA and mRNA data, our manuscript previous included these contents. In the section “ Predicted target genes and functional annotation ” from “ Methods ”, we have mentioned “According to previous differential expression genes (DEGs) between Meishan F18-resistant and -sensitive groups by transcriptome sequencing (data are available at NCBI’s SRA, PRJNA271310), we further selected potential target mRNAs related to E. coli F18 infection. On this basis, the regulatory network of DEMs and potential target mRNAs was established by Cytoscape software [24].”. In the section “ miRNA target gene prediction, GO enrichment and KEGG pathway analysis ” from “ Results ”, we have mentioned “Based on previous DGEs and functional enrichment, we further screened out important target genes related to E. coli F18 infection by Venny software (http://bioinfogp.cnb.csic.es/tools/venny/) and analyzed the regulatory network between important target genes and differential miRNAs (Fig. 7). Based on our results, we speculate that the α-(1, 2) fucosyltransferase 2 gene (FUT2) and Discs, large homolog 5 (DLG5) genes were the targets of down-regulated ssc-miR-218-3p, the MUC4 gene was the target of down-regulated ssc-miR-136, MyD88 was the target of up-regulated ssc-miR-499-5p, LBP and Toll-like receptor (TLR4) genes were the target of up-regulated ssc-miR-196b.”.